TL; DR: Ethereum will use at least ~99.95% less energy after the merger.
Ethereum will complete the transition to Proof of Stake in the coming months, which brings a host of improvements that have been theorized for years. But now that the Beacon chain has been up and running for a few months, we can dig into the numbers. One area we are excited to explore involves new estimates of energy use, as we finalize the process of spending a country’s energy by consensus.
There are no concrete statistics yet on power consumption (or even what hardware is used), so what follows is a rough estimate of future Ethereum power consumption.
Since a lot of people are running multiple validators, I’ve decided to use the number of unique addresses that made deposits as a proxy for how many servers there are today. Many takers could have used multiple eth1 addresses, but this largely defeats those with redundant setups.
At the time of writing, there are 140,592 validators for 16,405 unique addresses. This is obviously heavily skewed by exchanges and staking services, so removing them leaves 87,897 validators who are supposed to be staking from home. As a sanity check, this implies that the average home-staker runs 5.4 validators, which seems like a reasonable estimate to me.
power requirements
How much power does it take to run a beacon node (BN), 5.4 validation clients (VC) and an eth1 full node? Using my personal settings as a base, it’s around 15 watts. Joe Clapis (a Rocket Pool developer) recently ran 10 VCs, a Nimbus BN, and a Geth full node with a 10Ah USB battery bank for 10 hours, which means this setup averaged 5W. average staker is running such an optimized setup, so let’s call it 100W total.
Multiplying this with the 87k validators from before, means that home-stakers consume ~1.64 megawatts. Estimating the power consumed by the custodians is a bit more difficult, they run tens of thousands of validator clients with redundancy and backups.
To make life easier, let’s assume they use 100 W times 5.5 validators. Based on the staking infrastructure teams I’ve spoken to, this is a rough overestimate. The actual answer is something like 50 times less (and if you’re a custodial staking team consuming more than 5W/validator, contact me, I’m sure I can help).
In total, a proof-of-stake Ethereum consumes something on the order of 2.62 megawatts. This is not at the scale of countries, provinces, or even cities, but rather a small town (about 2,100 US households).
For reference, the Proof-of-Work (PoW) consensus on Ethereum currently consumes the power equivalent of a medium-sized country, but this is really necessary to keep a PoW chain secure. As the name suggests, PoW reaches a consensus based on which fork has the most “work” done. There are two ways to increase the rate of “work” being done, increase the efficiency of your mining hardware, and use more hardware at the same time. In order to prevent a chain from being successfully hacked, miners must be doing “work” at a higher rate than an attacker could. Since an attacker is likely to have similar hardware, miners must keep a lot of efficient hardware running to prevent an attacker from overtaking them, and all of this hardware consumes a lot of power.
Low PoW, as ETH price and hashrate are positively correlated. Therefore, as the price increases, in equilibrium so does the power consumed by the network. Under proof of stake, when the price of ETH increases, the security of the network increases as well (the value of ETH at stake is worth more), but the power requirements remain unchanged.
some comparisons
Digiconomist estimates that Ethereum miners currently consume 44.49 TWh per year, which equates to 5.13 gigawatts continuously. This means that PoS is ~2,000 times more energy efficient based on previous conservative estimates, reflecting at least a 99.95% reduction in total power usage.
If the energy consumption per transaction is greater than your speed, that’s ~35 Wh/tx (average ~60K gas/tx) or about 20 minutes of TV. By contrast, Ethereum PoW uses the energy equivalent of a house for 2.8 days per transaction and Bitcoin consumes 38 days’ worth of house.
Thinking in the future
While Ethereum continues to use PoW for now, that won’t be the case for much longer. In the last weeks, We have seen the appearance of the first test networks for the fusion, the name given to when Ethereum switched from PoW to PoS. Multiple teams of engineers are working overtime to ensure that the fusion arrive as soon as possible and without compromising security.
Scaling solutions (such as rollups and sharding) will help further decrease energy consumed per transaction by leveraging economies of scale.
Ethereum’s power-hungry days are numbered, and I hope that’s true for the rest of the industry as well.
Thanks to Joseph Schweitzer, Danny Ryan, Sacha Yves Saint-Leger, Dankrad Feist, and @phil_eth for their input.
